摘要
刀具剩余寿命预测对保证刀具加工质量有着重要意义,针对单一传感器信息预测精度低、抗干扰能力弱等问题,提出一种基于多源信息融合的刀具剩余寿命预测方法。首先,提取多传感器信号的时域、频域及时频域信息,构建刀具的多源信息矩阵;其次,计算特征序列与刀具寿命的斯皮尔曼系数,取系数大于阈值的信息为刀具的多源信息;最后,利用卷积门控循环神经网络(CNN-GRU)进行多源信息融合,实现刀具剩余寿命预测。在PHM2010刀具数据集实验验证,与LSTM、GRU相比,其预测精度分别提升了21%、22%,该结果证明了该方法具有更好的预测精度。
The prediction of tool residual life is very important to ensure the quality of tool processing.Aiming at the problems of low prediction accuracy and weak anti-interference ability of single sensor information,a method of tool residual life prediction based on multi-source information fusion was proposed.Firstly,the time domain,frequency domain and frequency domain information of multi-sensor signals are extracted,and the multi-source information matrix of the tool is constructed.Secondly,the spearman coefficients of feature sequence and tool life are calculated,and the information whose coefficients are greater than the threshold value is taken as the multi-source information of the tool.Finally,the convolution gated cyclic neural network(CNN-GRU)is used for multi-source information fusion to predict the tool remaining life.Compared with LSTM and GRU,the prediction accuracy of the proposed method is improved by 21%and 22%respectively in PHM2010 tool data set experiment,which proves that the proposed method has better prediction accuracy.
作者
申玉杰
孙显彬
刘伦明
曾实现
井陆阳
姜云春
SHEN Yu-jie;SUN Xian-bin;LIU Lun-ming;ZENG Shi-xian;JING Lu-yang;JIANG Yun-chun(School of Mechanical&Automotive Engineering,Qingdao University of Technology,Qingdao 266520,China;Qingdao Haier Industrial Intelligence Research Institute Co.,Ltd.,Qingdao 266000,China;不详)
出处
《组合机床与自动化加工技术》
北大核心
2022年第9期143-146,150,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
山东省自然科学基金(ZR2021ME026,ZR2020QE158)
山东省科技型中小企业创新能力提升工程项目(2021TSGC1045)
青岛市科技计划重点研发专项(21-38-04-0002)。
关键词
刀具
多源信息融合
卷积神经网络
门控循环单元
剩余寿命预测
tool wear
multi-source signal fusion
convolutional neural network
gating cycle unit
remaining life prediction
